Paper Reading | Latent Thought Models
LTM turns extra inference-time compute into explicit per-example posterior optimization over latent thought vectors, rather than more token-space decoding.
Posts
Paper Reading | Inverse Depth Scaling From Most Layers Being Similar
We know that deeper models generally have larger capacity, but what is the mechanism behind it? Given that the independent teacher regime is better aligned with the empirical signatures of real LLMs, ensemble averaging becomes the most plausible explanation for how real LLMs use depth.
A 1958 Physicist's Rules for Doing AI Research
An adapted version of Walther Bothe’s 1958 advice to young physicists, rewritten for modern AI research and archived from my original X article.
Reflections on My 2025 Fall Applications
A candid retrospective on my 2025 AI application cycle, including offers, mistakes, mentors, and the lessons I learned about research, planning, and building real connections.
How to Write a Good Letter of Recommendation?
A practical note on drafting stronger recommendation letters: be concrete, comparative, credible, and make it easier for busy professors to advocate for you.
Paper Reading | Cheating Popular LLM Benchmarks
Why popular LLM leaderboards can be gamed by structured outputs, how the cheating strategy works, and what this says about the reliability of automatic evaluation.